The PASTA project combines a longitudinal web-based survey with several smaller studies to gather objective data to complement the self-reported survey data (Fig. 1). An innovative web-based survey design deploys frequent follow-up questionnaires through a fully-automated platform. The survey is conducted in seven European cities: Antwerp BE, Barcelona ES, London GB, Oerebro SE, Rome IT, Vienna AT, Zurich CH (Fig. 2). Cutting edge technologies are used to collect objective (i.e. not self-reported) and relatively unbiased data in subsamples pulled from the survey respondents in some cities.
Longitudinal online survey
Survey contents
To align the survey contents with the research objectives, a detailed conceptual framework was developed based on previous work by others [23–25] and a scoping review of determinants of active mobility behaviour. Figure 3 illustrates the main domains of the framework. The framework distinguishes hierarchical levels for the various factors (i.e. city, individual, and trips), and three main domains or pathways that influence AM (and PA) behaviour, namely socio-geographical factors, socio-psychological factors, and rationale or mode choice related factors [23].
Data on contextual factors will be collected from publicly available GIS data and other data sources (i.e. weather data, population statistics, etc.) and by means of stakeholder interviews. Individual level data are being collected through the PASTA survey. Socio-psychological factors include concepts from extended theory of planned behaviour [24, 26], transtheoretical model [27] and a range of attitudinal questions. Various socio-geographical factors collected as objective city wide data are matched with corresponding questions in the survey to capture subjective perception of these same aspects (e.g. proximity of public transport). The main outcomes of interest in PASTA are travel behaviour (measured by frequency scale and 1-day travel diary adapted from KONTIV© design; and commuting route identification), physical activity behaviour (measured by PA single item [28] and the Global Physical Activity Questionnaire GPAQ with walking and cycling separated [29]), and traffic safety incidents (i.e. crashes and near misses – as used for example in the SHAPES project [16]). These are collected prospectively and measured repeatedly (see survey design and Fig. 1).
Survey design
To address several of the research questions, PASTA uses a longitudinal design, with a comprehensive baseline questionnaire and frequent short follow-ups.
The initial baseline questionnaire takes approximately 30 min and collects key socio-demographic, individual, household, health, attitudinal and other variables that identify the person and puts her/him into social context. Frequency of use of different modes and GPAQ questions gather information on mobility and PA habits. A one-day travel diary captures trips of the previous day in much detail. Thirteen days after completion of the last questionnaire, a short follow-up questionnaire which takes only about 5 min to complete is sent to the participant asking about PA and travel behaviour in the last seven days, and crashes and near misses since the last questionnaire (Fig. 1). Each third questionnaire is a somewhat longer follow-up including also a one-day travel diary, and taking about 10 min to complete. If a crash using AM is reported in one of the follow-ups, this prompts an additional crash questionnaire asking about crash circumstances, location, causes, injuries and other consequences.
Longitudinal designs are generally considered more challenging with respect to costs and complexity, and they put a higher burden on respondents. Several measures have been put in place to reduce attrition rates and ensure high data quality. The user-friendly and custom-made survey platform gives an intuitive overview of completed and open questionnaires. If a participant has not completed or finished a questionnaire, the platform is programmed to send them e-mail reminders after the 3rd, 10th and 20th day. The participant can log-in to the platform at any time and complete the unfinished questionnaire.
Preliminary data suggest drop-out rates of 20-25 % for the first short and for the long follow-up questionnaires; and approximately 15 % for the following short follow-up questionnaires.
Survey sample and recruitment
Based on a power-analysis, a pooled sample of 14000 respondents - 2000 in each of the seven partner cities - will be suited to address the key questions on AM and PA as well as to conduct in depth analyses of specific correlates and subgroups. Initial experiences and emerging issues are helping us in balancing the number of people recruited with the objectives of each analysis (and the number of correlates to be modelled). A number of exclusion criteria were applied: a minimum age of 16 years (18 years depending on the local ethical approval), and to be living and/or working/studying and/or regularly travelling in a PASTA city. On enrolment, participants register on the PASTA website and give informed consent (Additional file 1: Figure S1).
A standardized recruitment strategy was developed for all cities using an opportunistic approach. This included a press release after launching the survey platform; common promotional materials including postcards and leaflets; direct targeting of local stakeholders and community groups; and extensive use of social media. Within this framework there was room for local initiatives and targeted, city-specific recruitment. A coordinated effort was set-up at the beginning of the data collection and maintained throughout as we expect recruitment to continue for the duration of the study. To minimise attrition a user engagement strategy was developed, including incentivizing participation (lottery, except in Sweden where this is not allowed), regular contact with the respondents, branding of PASTA, posting on social media, and keeping the PASTA website up-to-date.
Technical implementation
The survey in the core module is implemented as an online web application, with a responsive design approach (i.e., the questionnaire can be completed across a wide range of devices – from mobile phones and tablets to desktop computers).
The PASTA platform is implemented in PHP with a PostgreSQL back-end database. In addition to the participant’s user interface, it also provides a researchers’ user interface and dashboard for real-time monitoring of recruitment and survey data collection, and a survey administration interface for survey creation and management (Additional file 1: Figures S1-4). All content was developed in English and translated into Swedish, Dutch, Catalan, Spanish, Italian, Swiss German, and Austrian German using the collaborative Pootle translation tool (http://pootle.translatehouse.org/). A formal testing protocol ensured systematic testing by the project partners before the official launch of the survey platform in November 2014. The survey can be accessed here: https://survey.pastaproject.eu/ (Additional file 2) and is planned to be online until October 2016.
Effectiveness of measures to promote active mobility
A key objective of the study is to evaluate the effectiveness of measures to promote AM with regards to their impacts on AM and PA. Other studies such as the iConnect study [25] in the UK have shown the effects of new physical infrastructures on AM and PA, using a similar approach. In comparison, the PASTA study investigates a wider range of AM measures defined here as actions or projects undertaken to increase the level of AM (in a specified population). For each city, the PASTA team is investigating one high-priority measure (hereafter ‘top measure’) to promote AM and PA (Additional file 1: Table S1). These measures range from infrastructure investments and built environment changes, such as bicycle racks and a dedicated cycling bridge, to soft measures such as workplace mobility management and individual marketing campaigns including ICT-elements.
Participants taking part in the longitudinal survey are identified as either being affected by the local top measure or being part of the ‘control group’, as defined by the distance between their home/work and the intervention address or by their response to a baseline questionnaire item asking about awareness and use of the relevant scheme (Fig. 4). These respondents are asked to complete a sequence of questionnaires ‘before’ (baseline and two short follow-up questionnaires) and ‘after’ the implementation of the top measure. After the implementation a re-entry and several follow-up questionnaires are sent to account for response lags (people do not change their behaviour immediately after an AM intervention). During the implementation period – referred to as the hibernation period – they do not receive any new questionnaires. A hibernation period was defined to avoid an effect of the implementation of the intervention itself (e.g. construction works) on the responses and to keep attrition rates as low as possible. For top measures that are implemented stepwise or have no clear time schedule, no hibernation phase was defined and respondents continue to receive follow-up questionnaires. Top measures will be evaluated over time (within subjects) and across study groups (i.e. affected versus control group). Power calculation indicated significant gains in power from repeated measurements before and after the implementation of the top measures, which are assumed to outweigh loss in respondents due to increased burden from repeated measurements.
Real-life study on route tracking & accelerometry
Given the known limitations of self-assessed levels of walking and cycling [30, 31], it is important to validate self-reported (subjective) data from the PASTA longitudinal online study. In a sample of volunteers selected from the longitudinal study (about 20 % of all respondents), the smart phone application Moves will be used to track journeys and automatically detect active travel modes over the study period (https://www.moves-app.com/), as shown previously by others [32, 33]. Using the participant’s PASTA identifier, data is downloaded from the Moves server and uploaded to the PASTA server (Additional file 1: Figure S5). Objective data on walking and cycling trips can again be compared to single questionnaire items like the mode frequency scale and the travel diary, or to physical activity summary variables like the GPAQ by computing Spearman correlation coefficients [31]. One goal is to derive correction factors for self-reported versus objective travel, which will be applied in the development of a HIA. Routes data from the Moves application will be used to analyse mode and route choice, and spatial aspects of AM.
In addition, in selected cities these same participants are also invited to wear an accelerometer during one week, providing objective data on overall PA. This will further allow validation of the adapted GPAQ questionnaire included in the PASTA survey by comparing the survey-reported values with accelerometer-derived levels of walking and cycling. We know from previous studies that time spent cycling may contribute to the disagreement between self-reported and objectively measured estimates of activity [31]. Therefore new accelerometry methods (triaxial, wrist-worn) are compared to standard accelerometry, especially to assess cycling which traditionally has been hard to monitor.
Because of the low user burden, no financial compensation is rewarded.
Real-life study on PA, air pollution & short-term health effects
We aim to collect objective data on possible health outcomes of AM. Air pollution exposure and the increased minute ventilation and dose during PA, is one of these potential negative effects. On the other hand, PA has a positive impact on health, maybe even on the same biomarkers. The study is designed as a repeated-measures study in a free-living population. In contrast to scripted studies, volunteers are tracked over a longer time period (one week) widening the window with prior information on air pollution exposure, PA levels, travel behaviour, and other confounders. Given this study design, mobile measurements of PA and exposure to TRAP are necessary; a number of previous studies managed to use new mobile sensors for this [34, 35]. By measuring air pollution and PA at the same time, the combined effects on health outcomes can be studied.
In three cities (Antwerp, Barcelona, London) a total of 120 participants are selected from the PASTA core study and equipped with tracking, air pollution and health sensors. Every participant carries all devices for one week while pursuing their regular activities (Additional file 1: Figure S6): a microAeth black carbon aerosol monitor (AethLabs, USA), a SenseWear (BodyMedia, USA) for PA, a Zephyr BioHarness (Zephyr, USA) for breathing rate and heart rate, a GPS (I-GOTU GT-600) and a smartphone (Samsung Galaxy SII, Korea) for geo-localization and accelerometry. This continuous 7-day measurement cycle is repeated in three contrasting seasons by every volunteer. At the beginning and at the end of the measurement week, volunteers visit the study centre and specific subclinical health biomarkers are measured in a controlled setting – these health markers can be linked to PA and TRAP exposure the hours and days before the measurement. The biomarkers were selected after a literature review, with non-invasive and direct read-out markers given preference. The final selection included both cardiovascular and respiratory parameters. Heart rate variability (HRV) is measured by the Zephyr BioHarness, and particle exposures have often been associated with lower HRV [8, 36–38], while blood pressure (measured by the Omron M10-IT, the Netherlands) is suspected to increase with increasing exposures to TRAP [38–40]. The microcirculation can be explored noninvasively by fundus images (Canon CR-2 Retinal Camera, Japan); the diameter of the vessels in the retina tends to decrease with an increase in TRAP [40, 41]. Exhaled NO is measured with the NIOX VERO (Aerocrine, Sweden); as a marker for lung inflammation exhaled NO generally increases after traveling and being exposed to TRAP [8, 9, 11, 36, 38, 42, 43]. Finally, a spirometry test is performed measuring lung function (EasyOne, ndd Medizintechnik AG, Switzerland). Multiple studies find no acute effects of exposure to TRAP or PA in healthy adults, others find small effects on some of the parameters [8, 9, 11, 36, 38, 43, 44]. Participants receive a small financial compensation for participation.
Prospective study on traffic safety incidents and investigation of crash locations
Objective and perceived safety are important barriers to AM [13, 17]. In the core longitudinal survey as well as in one of the add-on modules, traffic safety incidents are studied aiming to fill some of the research gaps identified. First of all, risks of having a traffic safety incident are poorly estimated up to now. Since travel behaviour and traffic safety incidents (crashes and near misses) are tracked over time, we can calculate exposure-adjusted (per km) risk estimates and further analyse these for relevant correlates, by incident categories, and by travel mode (walking, cycling, e-bikes). Such analyses are crucial for traffic safety improvement measures. Also, minor crashes and near misses are substantially underreported [45]. By specifically asking to report minor crashes and near misses in a prospective design (thus asking regularly), reporting levels should go up, and incidents not recorded in hospital, insurance or police records will be captured.
The role of built environment attributes, in particular infrastructure, is also poorly understood in this context. PASTA will conduct infrastructure audits of crash locations reported by pedestrians and cyclists, and compare these to randomly selected non-crash locations from respondents’ routes. Only a meticulous, exposure weighted case-crossover analysis can provide risk estimates for specific infrastructure types and attributes, as previously shown by Teschke et al. [46] and Vandenbulcke et al. [47].
Data analysis plan
The PASTA data analysis plan involves coordinated procedures for planning, organizing and documenting research, cleaning and preparing data, and analysing data. GitHub will be used as a practical tool for exchange of scripts (STATA or R), version control, issues and milestones between researchers (https://github.com/).
The collected data are automatically saved in a relational database management system (RDBMS) managed in one place. Data is anonymised before further analysis by stripping out or coding all information that would allow identification. The primary survey data is complemented with secondary data, such as GIS, meteorology, traffic, and land-use. A collection of indicators on AM measures, infrastructure, policies and relevant framework conditions is available for each city within PASTA, and can be used for qualitative data analyses.
Planned analyses cover a broad range of research questions, including building a predictor model for AM and on correlates of AM, PA and crash risk. The interrelation of AM and overall PA (‘substitution effects’) will be studied as one of the key outcomes. Qualitative data from interviews and workshops with stakeholders undertaken in parallel work in PASTA will be analysed and considered in the interpretation of findings and to the extent that it will be possible to build psychosocial constructs and latent variables, integrated in the statistical analysis and modelling. Separate analyses will focus on previously understudied topics such as e-bikes, bike-sharing and car-sharing programs. Pooled analyses comprising data from all seven cities will be preferred over single-city evaluations.
A specific objective of the PASTA project is to build a HIA on AM [19, 48], and to update the WHO Health Economic Assessment Tool (HEAT http://www.heatwalkingcycling.org/). This will help urban planners, transport planners and policy makers in the decision making process for investments in AM measures. Quantitative data from the longitudinal online survey and add-on experimental studies will provide HIA with an indication of a.o.:
-
▪The level of walking and cycling (average and distribution);
-
▪The contribution of walking and cycling to total PA;
-
▪Substitution of leisure time PA by AM;
-
▪The crash risk per kilometre walked/cycled in each city;
-
▪Air pollution exposure in different modes and its potential effects on health, and the combined effects of PA and air pollution on health.